ISR11 Relevance Feedback in an Information Retrieval System chapter W. Riddle T. Horwitz R. Dietz Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. v'-52 Results for Query No. 9 OA9AI[OCRerr]LYSIS Document Rank No. N&[OCRerr]e Relevant Correlation 1 39 51[OCRerr][OCRerr]S[OCRerr][OCRerr]RCH .43301270 2 47 l1l6RETRIEVA .42135049 3 11 l117FLAl[OCRerr]TING .40209035 4 62 809A [OCRerr]WCHANI .39900373 5 12 1113THE I[OCRerr]IE .34815531 6 2 814A [OCRerr] EFF .34299717 7 4 3L7ROLE OF / .32478490 8 17 3OIDOCUMENT .320750.5 9 9 206ANALYSIS .P$329423 10 23 815A FACET A .29329423 11 27 5041s RELEVA .28867513 12 21 1209SCIENCE .28724249 13 1 l104TI[OCRerr] IBM .28647316 i4 71 108sE[OCRerr]Ci[OCRerr][OCRerr]s .28517591 15 70 60IENC0D[OCRerr]TG .28306926 Ranking took .166657 seconds Relevant Document Ranks Document No. Name Relevant Correlation I33I[OCRerr]82[OCRerr]1 308IAkCiIIIr[OCRerr] R x .20701967 LLLLLI 4l8ENGLISH-L x .12038585 Normalized Recall = .5061728 Normalized Precision = .1717605 Rank Recall = .0361 Log Precision = .0936 I' = N TATEIGHT(I) * Document of Rank [OCRerr]ueryki) [OCRerr]uery[OCRerr]0) + Alpha 1:1 (i) I Document Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 i4 15 Weight -l -l 0 0 0 0 0 0 0 0 0 0 0 0 0 N = number of retrieved documents Alpha = 2 (d) Initial Results Use of Negative Weights Cosine Correlation Function Constant Alpha Strategy Figure 10 (continued)